Closed seas in eORCA025

[fmasson]

Dear colleagues,

Apologies if this issue has already been raised somewhere else - in that case, I'm happy to be redirected.

I have also submitted on the ShaCoNEMO (​http://forge.ipsl.jussieu.fr/shaconemo/ticket/52) but did not get responses so far.

Context

I'm running nemo_36_stable, revision 6396, with eORCA025 input configuration files from ShaCoNEMO. The grid consists in the extended ORCA025 grid, clipped (ncks -d y,157,1206) to get rid of the unnecessary land points in Antarctica.

I'm running with nn_closea = 0, that is, requesting that NEMO masks closed sea and lakes. The parameter jp_cfg is 025 in the namelist.

Analysis

There are three problems:

1. Great lakes (Canada/US border) and Lake Victoria are not masked
2. Black Sea is partly masked while it should not (it is connected to the Mediterranean)
3. Caspain Sea is incompletely masked while it should be fully covered.

See attached figure for an illustration.

Fix

I've noticed that the way closed seas and lakes are masked is hard-coded in routine dom_clo (in closea.F90). The following patch fixed the problem.

`
CASE ( 025 ) ! =========================

! ! clipped extended ORCA025
! ! =========================
ncsnr(1) = 1 ; ncstt(1) = 0 ! Caspian + Aral Seas
ncsi1(1) = 1330 ; ncsj1(1) = 675
ncsi2(1) = 1415 ; ncsj2(1) = 825
ncsir(1,1) = 1 ; ncsjr(1,1) = 1
!
ncsi1(2) = 777 ; ncsj1(2) = 710 ! Great lakes
ncsi2(2) = 848 ; ncsj2(2) = 768
!
ncsi1(3) = 1273 ; ncsj1(3) = 506 ! Lake Victoria
ncsi2(3) = 1295 ; ncsj2(3) = 536

`

[lovato]

Actually, in the shaconemo repository you can find an specific version of closea with a more flexible indexing that works both eORCA025 or clipped ORCA025

​http://forge.ipsl.jussieu.fr/shaconemo/browser/trunk/ORCA025_LIM3_PISCES/MY_SRC/closea.F90

I report here below the code changes of closea.f90 I refer to (same strategy used for eORCA1/ORCA1 grid)

            !                                        ! =======================
         CASE ( 025 )                                ! ORCA_R025 configuration
            !                                        ! =======================
            isrow = 1207 - jpjglo                    !  eORCA025 R025 - Using full isf­extended
                                                     !  domain for reference. - Adjust j­indices
            ncsnr(1)   = 1    ; ncstt(1)   = 0               ! Caspian + Aral sea
            ncsi1(1)   = 1330 ; ncsj1(1)   = 831 - isrow
            ncsi2(1)   = 1400 ; ncsj2(1)   = 981 - isrow
            ncsir(1,1) = 1    ; ncsjr(1,1) = 1
            !
            ncsnr(2)   = 1    ; ncstt(2)   = 0               ! Azov Sea
            ncsi1(2)   = 1284 ; ncsj1(2)   = 908 - isrow
            ncsi2(2)   = 1304 ; ncsj2(2)   = 933 - isrow
            ncsir(2,1) = 1    ; ncsjr(2,1) = 1
            !
            ncsnr(3)   = 1    ; ncstt(3)   = 0               ! Great Lakes
            ncsi1(3)   = 775  ; ncsj1(3)   = 866 - isrow
            ncsi2(3)   = 848  ; ncsj2(3)   = 931 - isrow
            ncsir(3,1) = 1    ; ncsjr(3,1) = 1
            !
            ncsnr(4)   = 1    ; ncstt(4)   = 0               ! Lake Victoria
            ncsi1(4)   = 1270 ; ncsj1(4)   = 661 - isrow
            ncsi2(4)   = 1295 ; ncsj2(4)   = 696 - isrow
            ncsir(4,1) = 1    ; ncsjr(4,1) = 1

I think the changes above should be added in the code nemo_v3_6_STABLE to get rid of this bug for the ORCA025 configuration.

Once this is agreed, we can close both tickets on nemo trac and shaconemo.

[timgraham]

Fixed in nemo_v3_6_STABLE at r7213.

I haven't fixed in the trunk for now because this code should be replaced as part of the simplification process at the next merge party. Just in case I'll change the version on the ticket to trunk and leave it open for now.